Life in the Earth system is not simply adapted to fixed or prescribed environmental conditions. The biosphere actually influences and controls, to a considerable extent, the setting for its own evolution. This uncontested insight is echoed by the much-debated ``Gaia Hypothesis'' (see, e.g., Lovelock), which speculates about self-stabilizing capabilities of the ``planetary super-organism''. Whether the hypothesis is correct or not, the underlying geophysiological approach to the explanation of geosphere-biosphere interactions is most valuable and may provide explanations for some puzzling observations. Long-term sedimentary records indicate, for instance, that there has always existed fluid water at the Earth's surface . This is a crucial hint to our planet's strong ability to self-regulate against external and internal driving forces. Although the solar luminosity was significantly lower in the past, the dead end of an ice planet has been avoided. A better understanding of this surprisingly resilient character of the coupled geosphere-biosphere system is highly relevant, especially, in view of the unintentional global experiments humankind is presently conducting via modification of the composition of the atmosphere or fragmentation of terrestrial vegetation cover.
A particularly useful ansatz for the investigation of geosphere-biosphere feedbacks is the Lovelock-Watson model (LWM) of ``Daisyworld'' [3, 4]. Despite its toy character, this model sheds much light on possible mechanisms of environment stabilization through evolutionary adaptation of terrestrial vegetation to varying insolation. Even more insights can be gained if the simple LWM is replaced by a 2D cellular automaton (CA) version , which takes into account a number of additional physical (e.g., lateral heat flow) and biological (e.g., food-web dynamics, competition and mutation) processes reflecting the interacting elements within the real Earth system.
Within our 2D model the area eligible for vegetation growth is initially a full square normalized to unity, i.e., a simply-connected domain. For the sake of realism, however, we also have to take into account that the area available for biospheric adaptation to ``Global Change'' forces is highly fragmented by civilisatory activities: urban settlements, infrastructures, agriculture, tourism, etc.. The implications of habitat fragmentation on biodiversity is at present a much-debated issue.
Our toy planet constitutes an ideal theatre for investigating this question and related ones in some depth; we specifically ask how the species make-up of the biosphere and the resulting self-stabilizing properties depend on landscape heterogeneity. The latter is mimic here in a well-defined way: we employ the percolation model from solid state physics  in order to mimic successive non-trivial reduction of growth space. Thus our paper is organized as follows: In Sect. 2 we describe in some detail the geophysiological simulation model used by introducing the pertinent elements to be considered one by one. In Sect. 3, we study the effects of percolation-type habitat fragmentation for different degrees of complexity of the biosphere make-up. The lessons to be learned from our results are reviewed in the short concluding section.